An Introduction to Hilbert-Huang Transform:

A Plea for Adaptive Data Analysis

Norden E. Huang
Research Center for Adaptive Data Analysis
National Central University


Data Processing and Data Analysis
Processing [proces < L. Processus < pp of
Procedere = Proceed: pro- forward + cedere, to
go] : A particular method of doing something.

Analysis [Gr. ana, up, throughout + lysis, a
loosing] : A separating of any whole into its parts,
especially with an examination of the parts to
find out their nature, proportion, function,
interrelationship etc.
Data Analysis
Why we do it?

How did we do it?

What should we do?
Why?
Why do we have to analyze data?
Data are the only connects we have with the reality;
data analysis is the only means we can find the truth
and deepen our understanding of the problems.
Ever since the advance of computer and
sensor technology, there is
an explosion of very complicate data.

The situation has changed from a thirsty for
data to that of drinking from a fire hydrant.
Henri Poincar

Science is built up of facts
*
,
as a house is built of stones;
but an accumulation of facts is no more a science
than a heap of stones is a house.

* Here facts are indeed our data.
Data and Data Analysis
Data Analysis is the key step in converting
the facts into the edifice of science.

It infuses meanings to the cold numbers,
and lets data telling their own stories and
singing their own songs.
Science vs. Philosophy
Data and Data Analysis are what separate
science from philosophy:

With data we are talking about sciences;
Without data we can only discuss philosophy.
Scientific Activities
Collecting, analyzing, synthesizing, and
theorizing are the core of scientific activities.

Theory without data to prove is just hypothesis.

Therefore, data analysis is a key link in this
continuous loop.
Data Analysis


Data analysis is too important to be left to
the mathematicians.

Why?!
Different Paradigms I
Mathematics vs. Science/Engineering
Mathematicians

Absolute proofs

Logic consistency

Mathematical rigor
Scientists/Engineers

Agreement with observations

Physical meaning

Working Approximations
Different Paradigms II
Mathematics vs. Science/Engineering
Mathematicians

Idealized Spaces

Perfect world in which
everything is known

Inconsistency in the different
spaces and the real world
Scientists/Engineers

Real Space

Real world in which knowledge is
incomplete and limited

Constancy in the real world within
allowable approximation
Rigor vs. Reality
As far as the laws of mathematics refer to
reality, they are not certain; and as far as
they are certain, they do not refer to reality.

Albert Einstein

How?
Data Processing vs. Analysis
All traditional data analysis methods are really for data
processing. They are either developed by or established
according to mathematicians rigorous rules. Most of the
methods consist of standard algorithms, which produce a
set of simple parameters.

They can only be qualified as data processing, not really
data analysis.

Data processing produces mathematical meaningful
parameters; data analysis reveals physical characteristics
of the underlying processes.
Data Processing vs. Analysis
In pursue of mathematic rigor and certainty,
however, we lost sight of physics and are forced
to idealize, but also deviate from, the reality.

As a result, we are forced to live in a pseudo-real
world, in which all processes are

Linear and Stationary


Trimming the foot to fit the shoe.
Available Data Analysis Methods
for Nonstationary (but Linear) time series
Spectrogram
Wavelet Analysis
Wigner-Ville Distributions
Empirical Orthogonal Functions aka Singular Spectral
Analysis
Moving means
Successive differentiations

Available Data Analysis Methods
for Nonlinear (but Stationary and Deterministic)
time series
Phase space method
Delay reconstruction and embedding
Poincar surface of section
Self-similarity, attractor geometry & fractals

Nonlinear Prediction

Lyapunov Exponents for stability

Typical Apologia
Assuming the process is stationary .

Assuming the process is locally stationary .

As the nonlinearity is weak, we can use perturbation
approach .

Though we can assume all we want, but
the reality cannot be bent by the assumptions.
The Real World


Mathematics are well and good but nature
keeps dragging us around by the nose.

Albert Einstein
Motivations for alternatives:
Problems for Traditional Methods
Physical processes are mostly nonstationary

Physical Processes are mostly nonlinear

Data from observations are invariably too short

Physical processes are mostly non-repeatable.

Ensemble mean impossible, and temporal mean might not
be meaningful for lack of stationarity and ergodicity.

Traditional methods are inadequate.
What?
The job of a scientist is to listen carefully to
nature, not to tell nature how to behave.

Richard Feynman


To listen is to use adaptive methods and let the data sing, and
not to force the data to fit preconceived modes.

The Job of a Scientist
How to define nonlinearity?
Based on Linear Algebra: nonlinearity is
defined based on input vs. output.

But in reality, such an approach is not
practical. The alternative is to define
nonlinearity based on data characteristics.
Characteristics of Data from
Nonlinear Processes
( )
3
2
2
2
2
2
d x
x cos t
dt
d x
x cos t
dt
Spring with positiondependent cons tant ,
int ra wave frequency mod ulation;
therefore, we need ins tan
x
1
taneous frequenc
x
y.
c e
c e
+ + =
+ =

Duffing Pendulum
2
2
2
( co . ) s 1
d x
x t x
dt
c e = + +
x
( )
p
2 2 1 / 2 1
i ( t )
For any x( t ) L ,
1 x( )
y( t ) d ,
t
then, x( t )and y( t ) formthe analytic pairs:
z( t ) x( t ) i y( t ) ,
where
y( t )
a( t ) x y and ( t ) tan .
x( t )
a( t ) e
t
u
t
t
t t
u

e
=

= + =
= + =
}
Hilbert Transform : Definition
Hilbert Transform Fit
Conformation to reality rather then to
Mathematics

We do not have to apologize, we should use
methods that can analyze data generated by
nonlinear and nonstationary processes.

That means we have to deal with the intrawave
frequency modulations, intermittencies, and
finite rate of irregular drifts. Any method
satisfies this call will have to be adaptive.
The Traditional Approach of
Hilbert Transform for Data Analysis
Traditional Approach
a la Hahn (1995) : Data LOD
Traditional Approach
a la Hahn (1995) : Hilbert
Traditional Approach
a la Hahn (1995) : Phase Angle
Traditional Approach
a la Hahn (1995) : Phase Angle Details
Traditional Approach
a la Hahn (1995) : Frequency
Why the traditional approach
does not work?
Hilbert Transform a cos + b : Data
Hilbert Transform a cos + b :
Phase Diagram
Hilbert Transform a cos + b :
Phase Angle Details
Hilbert Transform a cos + b :
Frequency
The Empirical Mode Decomposition
Method and Hilbert Spectral Analysis

Sifting
Empirical Mode Decomposition:
Methodology : Test Data
Empirical Mode Decomposition:
Methodology : data and m1
Empirical Mode Decomposition:
Methodology : data & h1
Empirical Mode Decomposition:
Methodology : h1 & m2
Empirical Mode Decomposition:
Methodology : h3 & m4
Empirical Mode Decomposition:
Methodology : h4 & m5
Empirical Mode Decomposition
Sifting : to get one IMF component
1 1
1 2 2
k 1 k k
k 1
x( t ) m h ,
h m h ,
.....
.....
h m h
. h c
.

=
=
=
=
Two Stoppage Criteria : S and SD

A. The S number : S is defined as the consecutive
number of siftings, in which the numbers of zero-
crossing and extrema are the same for these S siftings.
B. SD is small than a pre-set value, where
T
2
k 1 k
t 0
T
2
k 1
t 0
h ( t ) h ( t )
SD
h ( t )

=
Empirical Mode Decomposition:
Methodology : IMF c1
Definition of
the Intrinsic Mode Function (IMF)
Any functionhaving the same numbers of
zero cros sin gs and extrema,and alsohaving
symmetric envelopes defined by local max ima
and minima respectively is defined as an
Intrinsic Mode Function( IMF ).
All IMF enjoys good Hilbert Transfo

i ( t )
rm :
c( t ) a( t )e
u
=

Empirical Mode Decomposition
Sifting : to get all the IMF components
n
j n
j
1 1
n
1
1 2 2
1 n n
x( t ) c r ,
r c r ,
.......
r
x( t ) c r
c r .
.
=

=
=
=
=

1 1
1 2 2
n 1 n n
n
j n
j 1
x( t ) c r ,
r c r ,
x( t ) c r
. . .
r c r .
.

=
=
=
=
=
Empirical Mode Decomposition:
Methodology : data & r1
Empirical Mode Decomposition:
Methodology : data and m1
Empirical Mode Decomposition:
Methodology : data, r1 and m1
Empirical Mode Decomposition:
Methodology : IMFs
Definition of Instantaneous Frequency
i ( t )
t
The Fourier Transformof the Instrinsic Mode
Funnction, c( t ), gives
W( ) a( t ) e dt
By Stationary phase approximationwe have
d ( t )
,
dt
This is defined as the Ins tantaneous Frequency.
u e
e
u
e

=
=
}
Definition of Frequency
Given the period of a wave as T ; the frequency is
defined as



1
.
T
Equivalence :
The definition of frequency is equivalent to
defining velocity as

Velocity = Distance / Time

Instantaneous Frequency
distance
Velocity ; mean velocity
time
dx
Newton v
dt
1
Frequency ; mean frequency
period
d
HH
So that both v and
T defines the p
can appear in differential equations.
hase function
dt
e
u
e
=
=
=
=
The combination of Hilbert Spectral Analysis and
Empirical Mode Decomposition is designated as
HHT
(HHT vs. FFT)
Jean-Baptiste-Joseph Fourier
1807 On the Propagation of Heat in Solid Bodies
1812 Grand Prize of Paris Institute
Thorie analytique de la chaleur
... the manner in which the author arrives at
these equations is not exempt of difficulties and
that his analysis to integrate them still leaves
something to be desired on the score of generality
and even rigor.
1817 Elected to Acadmie des Sciences
1822 Appointed as Secretary of Math Section
paper published
Fouriers work is a great mathematical poem.
Lord Kelvin
Comparison between FFT and HHT
j
j
t
i t
j
j
i ( )d
j
j
1. FFT :
x( t ) a e .
2. HHT :
x( t ) a ( t ) e .
e
e t t
= 9
}
= 9

Comparisons:
Fourier, Hilbert & Wavelet
An Example of Sifting
Length Of Day Data
LOD : IMF
Orthogonality Check
Pair-wise %

0.0003
0.0001
0.0215
0.0117
0.0022
0.0031
0.0026
0.0083
0.0042
0.0369
0.0400
Overall %

0.0452
LOD : Data & c12
LOD : Data & Sum c11-12
LOD : Data & sum c10-12
LOD : Data & c9 - 12
LOD : Data & c8 - 12
LOD : Detailed Data and Sum c8-c12
LOD : Data & c7 - 12
LOD : Detail Data and Sum IMF c7-c12
LOD : Difference Data sum all IMFs
Traditional View
a la Hahn (1995) : Hilbert
Mean Annual Cycle & Envelope: 9 CEI
Cases
Mean Hilbert Spectrum : All CEs
Tidal Machine
Properties of EMD Basis
The Adaptive Basis based on and derived from
the data by the empirical method satisfy nearly
all the traditional requirements for basis
a posteriori:
Complete
Convergent
Orthogonal
Unique
Hilberts View on
Nonlinear Data
Duffing Type Wave
Data: x = cos(wt+0.3 sin2wt)
Duffing Type Wave
Perturbation Expansion
( )
( ) ( )
For 1 , we can have
x( t ) cos t sin2 t
cos t cos sin2 t sin t sin sin2 t
cos t sin t sin2 t ....
1 cos t cos 3 t ....
2 2
This is very similar tothe solutionof Duffing equation .
c
e c e
e c e e c e
e c e e
c c
e e
= +
=
= +
| |
= + +
|
\ .
Duffing Type Wave
Wavelet Spectrum
Duffing Type Wave
Hilbert Spectrum
Duffing Type Wave
Marginal Spectra
Duffing Equation
2
3
2
.
Solved with for t 0 to 200 with
1
0.1
od
0.04 Hz
Initial condition :
[ x( o ) ,
d x
x x c
x'( 0 ) ] [ 1
os t
, 1]
3
t
e2
d
tb
c
c e

e
=
=
=
=
=
+ + =
Duffing Equation : Data
Duffing Equation : IMFs
Duffing Equation : Hilbert Spectrum
Duffing Equation : Detailed Hilbert Spectrum
Duffing Equation : Wavelet Spectrum
Duffing Equation : Hilbert & Wavelet Spectra
Speech Analysis
Nonlinear and nonstationary data
Speech Analysis
Hello : Data
Four comparsions D
Global Temperature Anomaly
Annual Data from 1856 to 2003
Global Temperature Anomaly 1856 to 2003
IMF Mean of 10 Sifts : CC(1000, I)
Statistical Significance Test
Data and Trend C6
Rate of Change Overall Trends : EMD and Linear
What This Means
Instantaneous Frequency offers a total different
view for nonlinear data: instantaneous
frequency with no need for harmonics and
unlimited by uncertainty.

Adaptive basis is indispensable for
nonstationary and nonlinear data analysis

HHT establishes a new paradigm of data
analysis
Comparisons
Fourier Wavelet Hilbert
Basis a priori a priori Adaptive
Frequency Convolution:
Global
Convolution:
Regional
Differentiation:
Local
Presentation Energy-frequency Energy-time-
frequency
Energy-time-
frequency
Nonlinear no no yes
Non-stationary no yes yes
Uncertainty yes yes no
Harmonics yes yes no
Conclusion
Adaptive method is the only scientifically
meaningful way to analyze data.

It is the only way to find out the underlying
physical processes; therefore, it is
indispensable in scientific research.

It is physical, direct, and simple.
History of HHT

1998: The Empirical Mode Decomposition Method and the Hilbert Spectrum for
Non-stationary Time Series Analysis, Proc. Roy. Soc. London, A454, 903-995.
The invention of the basic method of EMD, and Hilbert transform for determining
the Instantaneous Frequency and energy.
1999: A New View of Nonlinear Water Waves The Hilbert Spectrum, Ann. Rev.
Fluid Mech. 31, 417-457.
Introduction of the intermittence in decomposition.
2003: A confidence Limit for the Empirical mode decomposition and the Hilbert
spectral analysis, Proc. of Roy. Soc. London, A459, 2317-2345.
Establishment of a confidence limit without the ergodic assumption.
2004: A Study of the Characteristics of White Noise Using the Empirical Mode
Decomposition Method, Proc. Roy. Soc. London, (in press)
Defined statistical significance and predictability.
2004: On the Instantaneous Frequency, Proc. Roy. Soc. London, (Under review)
Removal of the limitations posted by Bedrosian and Nuttall theorems for
instantaneous Frequency computations.
Current Applications

Non-destructive Evaluation for Structural Health Monitoring
(DOT, NSWC, and DFRC/NASA, KSC/NASA Shuttle)
Vibration, speech, and acoustic signal analyses
(FBI, MIT, and DARPA)
Earthquake Engineering
(DOT)
Bio-medical applications
(Harvard, UCSD, Johns Hopkins)
Global Primary Productivity Evolution map from LandSat data
(NASA Goddard, NOAA)
Cosmological Gravity Wave
(NASA Goddard)
Financial market data analysis
(NCU)
Advances in Adaptive data Analysis:
Theory and Applications
A new journal to be published by
the World Scientific


Under the joint Co-Editor-in-Chief
Norden E. Huang, RCADA NCU
Thomas Yizhao Hou, CALTECH
in the January 2008

Oliver Heaviside
1850 - 1925

Why should I refuse a good dinner
simply because I don't understand
the digestive processes involved.